This year, we completed a four year project involving whole genome sequencing without selective pressure to map the locations of over 43,000 single base mutations generated in diploid, mismatch repair (MMR) proficient and MMR deficient budding yeast strains encoding wild type or variants of the three major nuclear DNA replicases. We found that on average, the nuclear DNA replication machinery generates less than one mismatch per genome, and in combination with MMR achieves a genome-wide per base error rate of &#8805; 1.7 x 10-10. In the absence of MMR and purifying selection, replication error patterns strongly depend on the location of replication origins and the local DNA sequence, the latter seen as preferred sequences for single base substitutions and deletions. Error rates also vary during early versus late replication, in linker versus nucleosome-bound DNA and in coding versus non-coding DNA. This genome-wide view reveals that replication fidelity is amazingly high yet heterogeneous, in patterns that suggest the underlying mechanisms by which replication influences genome stability and composition and vise versa.